JP4656727B2 - Active vibration isolation system identification with improved noise reduction - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to active control systems that reduce ambient vibrations, and more particularly, to using signal filtering to generate control signals that reflect adjustments to estimated vehicle structural dynamics. This control signal is sent to one or more vibration generators that reduce ambient vibrations of the vehicle.
[0002]
[Prior art]
In a vehicle such as a helicopter or an airplane, vibration may reach an undesirable level, resulting in an uncomfortable environment inside the vehicle. This is especially true for helicopters that vibrate in response to loads generated by the moving blades as they move and support the aircraft through the air. Traditionally, helicopter manufacturers use a variety of vibration isolation methods that use the principles of mechanical isolation and absorption in order to be able to accept the comfort of an aircraft. These methods may be described as “passive” and attack vibration by modifying the aircraft's inherent structural dynamics and “detuning” its response to common frequencies in noisy load characteristics. To do. Another method uses a powered actuator to apply a vibration load to the structure in such a way as to generate a vibration field that negates ambient vibrations. This type of method is represented as “active” in the anti-vibration region in the sense that it is commanded to actively generate a vibration field that the actuator disables.
[0003]
The key component to successfully implement an active vibration isolation system is to appropriately determine the command to the actuator. If the relationship between the actuator forces and the vibration response to these forces is known in advance, a “deterministic” control algorithm can be used to properly select the commands to the actuators. However, structural dynamics are affected by fuel consumption, cargo redistribution, or in the case of helicopters because the rotor load frequency changes when the rotor speed changes during operation. Thus, the relationship between actuator force and vibration response is often not known in advance or changes. In such cases, “online system identification” may be used to establish or modify an estimate of the relationship between actuator force and vibration response using information obtained while changing control. . This estimate is used during subsequent control iterations to reduce vibrations.
[0004]
A typical environment for a control system in which the present invention is suitably used is described in "Improvement and Evaluation of Helicopter Real-Time Self-Adaptive Active Vibration Isolator Algorithm" on pages 1-30 of NASA Contractor Report 3821. Is the smallest distributed control device environment. ThisThe controller uses an approximation with a quasi-static linear relationship that can be represented by the following transfer matrix relationship between the actuator command and the system response.
[0005]
[Expression 2]
Z_i= Τ (U_i-U_i-1) + Z_i-1+ V + E
Where Z_iIs the n-dimensional measured vibration response at time i, Z_i-1Is the n-dimensional state vector of the vibration response at time i−1, U_iIs the m-dimensional command for the actuator at time i, U_i-1Is the m-dimensional command for the actuator at time il, V is the n-dimensional change in the measured vibration due to the disturbance, E is the change in the measured vibration due to the change in measurement noise, and τ is the current estimate of n by the m-dimensional transfer matrix Value.
[0006]
The minimal distributed controller is obtained by minimizing the following performance categories:
[0007]
[Equation 3]
J = E {(Z_i-Z_opt)^TW_Z(Z_i-Z_opt) + U_i ^TW_UU_i+ ΔU_i ^TW_{Δ U}ΔU_i} Where J is a figure of merit (scalar), E is the expected value, Z_optIs the desired vibration vector at the sensor position (typically zero), Z is the measured vibration vector at the sensor position, W_ZIs the diagonal weighting matrix on the output (vibration) parameters, U is the command input to the actuator, W_UIs a diagonal weighting matrix that constrains the amplitude of the actuator command, ΔU is the change in command input to the actuator, W_{Δ U}Is a diagonal weighting matrix that constrains the rate of change of the command input, T is a vector or matrix transpose, and i is a discrete time increment counter.
[0008]
Since the figure of merit J includes measured output parameters and manipulated variables, each output parameter and manipulated variable can be individually weighted to be more important or less important than the other components.
[0009]
When a deterministic control device is used for the local model, the optimal change ΔU of the command input to the actuator for the rotation of the i-th rotor blade is as follows.
[0010]
[Expression 4]
ΔU_i= -D [W_UU_i-1+ Τ^TW_Z(Z_i-1-Z_opt]]
Where ΔU_iIs the optimum manipulated variable required to minimize the figure of merit, D = (τ^TW_Zτ + W_U+ W_{Δ U})^-1And the other variables are the same as described above.
[0011]
When J is minimized, the solution for the optimal controller provides information to determine the next control step. In order to improve the control performance, an accurate estimate of the matrix τ is very important. In real-time applications, it is useful to use a Kalman filter to track the elements of τ. Previous attempts to accurately track the value τ failed to adequately suppress the adverse effects of system and measurement noise due to the Kalman filter. For example, King US Pat. No. 4,819,182 discloses a method and apparatus for reducing helicopter fuselage vibration using a plurality of actuators that vibrate at a frequency corresponding to the excitation frequency. A plurality of accelerometers generate a signal indicative of dynamic acceleration. The actuator is controlled by the processing device. This patent does not disclose modifying the covariance value as a means to evaluate system dynamics.
[0012]
[Problems to be solved by the invention]
In many applications, the implementation of the Kalman filter algorithm assumes that the measurement noise and system noise are independent of each other, have a Gaussian distribution, and have a zero mean. If these assumptions are valid, the standard Kalman filter method can be an effective way to reduce the negative effects of noise on the estimation of the relationship between actuator force and vibration. If these assumptions are not valid, the standard method is not valid and results in an inferior estimate, thus reducing the performance of the control system that uses the estimate for control decisions. The purpose of the present invention is to refer to the disadvantages of the standard Kalman filter method in this respect.
[0013]
[Means for Solving the Problems]
It is an object of the present invention to provide a control system that utilizes dynamically validated and updated data to generate control signals that are used to reduce ambient vibrations. Accordingly, one embodiment of the present invention is directed to a system having at least one sensor attached to a vehicle to generate a continuous first and second ambient vibration signal. These ambient vibration signals correspond to the first and second detection states. One or more actuators are attached to the vehicle to generate vibration reducing actuator forces that are applied to the vehicle. A data module is coupled to the sensors and actuators for storing data. This data relates to at least one actuator force with at least one sensor response to that force and outputs accumulated data. A control circuit including a Kalman filter receives continuous first and second ambient vibration signals from the sensor and receives output data from the data module. The control circuit generates a first control signal and sends it to the associated actuator corresponding to the first ambient vibration signal. The control circuit utilizes Kalman filtration and covariance, which is a function of actuator force that modifies data in the data module. The control circuit outputs a second control signal that is a function of the second ambient vibration signal and the correction data of the data module.
[0014]
The second embodiment of the present invention relates to a system for reducing noise in a vehicle. The system includes sensing ambient vibrations in the vehicle. An ambient vibration signal corresponding to the detected ambient vibration is generated. This ambient vibration signal is processed to generate a control signal. The control signal is a function of dynamic covariance corresponding to a constant value divided by a change in actuator force and raised to an integer power. The control signal is an output to an actuator that generates a vibration field to reduce ambient vibration. The actuator force is a function of the control signal sent from the control circuit.
[0015]
【Example】
The present invention provides on-line system identification with increased vibration reduction by generating a control signal that is used to operate a vibration generator to reduce vibrations in the vehicle. This control signal is generated by a control circuit that uses an estimate of the vehicle's structural dynamics as a reference for calculating an optimal actuator command. An estimate of the vehicle's structural dynamics is determined by a modified Kalman filter method that uses the response information collected during application of the actuator command in the preceding control step. The modified Kalman filter method can more accurately assess the structural dynamics of an actual vehicle in the presence of noise than the standard Kalman filter method. As a result of improved estimation, the control circuit is more effective in reducing vibration.
[0016]
Reference is made to the drawings. In the several figures, similar reference signs identify corresponding or similar components. FIG. 1 illustrates an improved vibration reduction system 10 that may be utilized with a vehicle 122. The vehicle 122 includes sensors 110 (a) to (d) that detect vibrations and send a signal indicating the detected vibrations to the control device 118. The control device 118 processes the received vibration signal, generates a control command signal, and sends it to the actuators 114 (a) and 114 (b). Actuator 114 responds to the control command by generating a compensating force to reduce detected vibrations in vehicle 122.
[0017]
Although the vehicle 122 is depicted as a helicopter in FIG. 1, a fixed wing aircraft is also suitable, or may be a flight simulator device or any device that generates ambient vibrations to be detected from a vibration source. The vehicle 122 includes one or more sensors 110 attached to the vehicle 122. Illustratively, the sensors 110 are shown as four sensors 110 (a)-(d), but may be any number of sensors suitable for the vehicle 122 design. The number and location of sensors 110 is a design choice and is not critical to understanding the present invention. FIG. 1 shows sensors 110 (a) and 110 (b) attached to the cockpit of the vehicle 122. Sensor 110 (c) is located in the front passenger compartment of vehicle 122, and sensor 110 (d) is located in the rear passenger compartment. The sensor 110 detects ambient vibration in the vehicle 122. The ambient vibration detected by the sensor 110 is generated by a vibration load. Examples of the vibration load include vibration generated from the rotor blade 126 and vibration generated from the stabilizer 130. Also, external forces such as turbulence, downwash, gusts and other vibration generating events cause unwanted ambient vibrations on the vehicle 122.
[0018]
The sensor 110 detects the vibration data and sends a corresponding ambient vibration signal to a control circuit (also referred to as a control device) 118. The control circuit 118 processes the ambient vibration signal received from the sensor 110 via the intermediate connector 136. The intermediate connector 136 may be any device capable of sending signals from the sensors 110 (a)-(d) to the control circuit 118. A typical intermediate connector 136 is a wire or infrared signal transmission means.
[0019]
The control circuit 118 includes memory and processing capabilities and is typically a processing unit such as a digital signal processing unit (DSP), a microprocessor or an 80486 processor. Those skilled in the art will recognize that virtually any processor with appropriate memory and speed can be used for processing performed by the control circuit 118. The control circuit 118 executes software or is preprogrammed to filter the sensed ambient vibration signal received from the sensor 110. In particular, control circuit 118 utilizes Kalman filtering capabilities and software or hardware to generate covariance values that validate, update, and utilize data in system 10. The control circuit 118 embodies any control mechanism that attempts to minimize some mathematical functions of the system state and inputs. These systems are known to those skilled in the art as “optimal control systems”. The control circuit 118 generates one or more control signals (shown as a single control signal for purposes of illustration) that are sent to the vibration generator unit 114 shown as 114 (a) and (b). This control signal is the result of a confirmation procedure that determines the validity of the detected ambient vibration signal received from the sensor 110.
[0020]
A vibration generator unit (herein referred to as generating unit or actuator) 114 is suitably coupled to control circuit 118 via intermediate connector 138 and receives one or more control signals from control circuit 118. The intermediate connector 138 may be a wire, an infrared transmitting means as known to those skilled in the art, or any means that sends a control signal from the control circuit 118 to the vibration generator unit 114. Although FIG. 1 shows vibration generators 114 (a) and (b), any number of generator units suitable for the vehicle 122 design can be used. FIG. 1 shows two generator units 114 (a) and (b) for illustrative purposes only, and the quantity and location of the generator units 114 is a design choice and for understanding the present invention. Not serious. The generator unit 114 can be an actuator or any device capable of generating a vibration field. The generator unit 114 is typically mounted near the main rotor 126 when the vehicle is a helicopter, and is mounted at an appropriate position when the vehicle is another vehicle. The generator units 114 (a) and (b) are attached to the vehicle 122 and effectively generate a vibration reduction field. This vibration reduction field is used to counteract the vibrations generated from the aforementioned vibration load sources that adversely affect the dynamics of the vehicle 122.
[0021]
FIG. 2 shows a block diagram of the system 10 that includes a control circuit 118 coupled to the sensor unit 110 and the vibration generator unit 114. The sensor unit 110 generates a vibration signal corresponding to ambient vibration detected in the environment, typically inside a vehicle. This signal is sent to the control circuit 118 via the intermediate connector 136. The control circuit 118 includes a processor, shown as 148, one or more monitoring unit 144, such as a CRT, a memory 146 storing a plurality of modules, typically data structures, and one or more vehicle dynamics. It is appropriate to consist of a combination with the data matrix 147. The control circuit 118 includes one or more Kalman filters 140 (a) (i). The control circuit 118 includes an algorithm and software or hardware instructions, and inverts the vibration signal received from the sensor 110 or processes the received signal to generate an inverted vibration signal. Unit 114 generates a vibration reduction field. The processing commands for the control device 118 are appropriately stored in the memory 146 of the control circuit 118 or downloaded to the control circuit 118 as software. The memory 146 also stores information related to vehicle dynamics. This information is typically stored as a data matrix 147 containing data indicating an estimate of the vehicle's response based on previous vibration generator forces. Alternatively, the contents of this data matrix may be stored in a memory arranged away from the control circuit 118. The control circuit 118 includes one or more Kalman filters 140 (a)... (I) (where i is any number according to the design of the system, but in this specification a plurality of Kalman filters 140 (a). Described as a single Kalman filter 140).
[0022]
The Kalman filter 140 updates the state estimate based on the previous state estimate and the input of observed or measured system variables. These system variables may be measured in real time so that system variable data can be input directly into the Kalman filter 140, or may be accumulated before the system variable data is input into the Kalman filter 140.
[0023]
The controller algorithm, suitably stored or pre-programmed in the control circuit 118, processes the signal received from the sensor 110 and generates an appropriate control signal that is sent to the vibration generator unit 114 via the intermediate connector 138. Mechanism.
[0024]
Ambient vibration measurements measured by the sensor 110 over a period of time may be collected in an nx 1-dimensional vector denoted as Z. Similarly, the force exerted by the vibration generator (actuator) 114 may be collected in an mx1-dimensional vector denoted by U. The relationship between the change in actuator force (ΔU) and the resulting change in sensor measurement (ΔZ) may be expressed as the matrix equation ΔZ = τΔU. Here, τ is an nxm-dimensional data matrix 147 (values stored in the data matrix τ 147 are appropriately used in the above-described equation). The values of the elements of the data matrix τ 147 are determined by physical properties, ie the vehicle's structural dynamics, and are stored appropriately in the matrix 147. This matrix 147 is shown in the control circuit 118, but may be suitably arranged away from the control circuit 118 and communicated with the control circuit 118. Examples of parameters of the data matrix τ are as follows. τ₁₁Is the change in the response at the sensor number 1 due to the unit change of the force command at the actuator number 1,₁₂Is a change in response in the sensor number 1 due to a unit change in force command in the actuator number 2.
[0025]
Typically, the magnitude of ΔU and ΔZ indicates a measure of a quantity that is the result of manipulating the actual quantity to produce a function of that quantity (eg, electrical signal amplification and attenuation or signal processing). . Many algorithms may be used to determine control based on this model. One example of such an algorithm that attempts to minimize the scalar figure of merit is:
[0026]
[Equation 5]
J_cost= Z^TW_ZZ + U^TW_UU + ΔU^TW_{Δ U}ΔU
Where W_ZIs the diagonal weighting matrix of sensor measurements, W_UIs a diagonal weighting matrix that constrains the manipulated variable, W_{Δ U}Is a diagonal weighting matrix that constrains the change rate of the manipulated variable, ΔU is a change in actuator force, U is an actuator command force, Z is an ambient vibration signal, and T is a transposed matrix of a vector or matrix.
[0027]
Using the above objectives in the context of discrete time control applications, the control algorithm for the i th step takes the form:
[0028]
[Formula 6]
ΔU_i=-(Τ^TW_Zτ + W_U+ W_{Δ U})^-1[W_UU_i-1+ Τ^TW_Z(Z_i-1]]
Here, (−1) indicates matrix inversion, and τ is the above-described transfer matrix relating to sensor response change with respect to actuator command change.
[0029]
A very important component of the performance of this control algorithm is the accurate identification of the data matrix τ 147 in which the elements are stored, ie the determination of the value of τ that appropriately characterizes the structural dynamics of the vehicle. One way to do this identification before operating the control circuit 118 is to simply command one actuator force at a time from each actuator 114 while measuring the response at each sensor 110. However, in the conventional ambient vibration reduction system, this is difficult when the elements of the data matrix τ147 are non-constant, that is, when the elements of τ change during operation. In the case of a vehicle active vibration control system such as a helicopter, many factors cause non-constant elements of the data matrix τ 147. For example, when the rotor blade speed changes and the frequency changes, the sensitivity of the sensor 110 response due to the actuator force changes due to a change in the relationship between the excitation frequency and the natural frequency of the helicopter fuselage. Some other factors that can produce changes in the elements of the data matrix τ 147 are changes in the fuselage mass distribution resulting from passenger movement or fuel consumption, which affects the structural dynamics of the helicopter.
[0030]
If the values of the elements of the data matrix τ 147 used in the control circuit 118 do not accurately reflect the characteristics of the vehicle, the performance of the control circuit 118 can be significantly degraded to some point of instability. In order to more accurately determine the value of τ, which is an element of the data matrix τ147, the Kalman filter can be used in the following form.
[0031]
[Expression 7]
τ_i= [Τ^T _i-1+ K_i(ΔZ_i ^T-ΔU_i ^Tτ^T _i-1]]^T
Where K_iIs the Kalman gain matrix, K_i= P_iΔU_iR_i ^-1, M_i= P_i-1+ Q_i-1, P_i= M_i-M_iΔU_i(ΔU_i ^TM_iΔU_i+ M_i)^-1ΔU_i ^TM_i, ΔU is a change in actuator force, ΔZ is a change in sensor signal, T is a transpose matrix, R is a covariance of vibration measurement noise, Q is a covariance of system noise, ie a change in operating environment or system structure dynamics This is the covariance of fluctuations in vibration due to changes.
[0032]
The Kalman filter reflects changes in the vehicle's structural dynamics in the estimate of τ while filtering out the effects of measurement and system noise.
[0033]
  In the derivation of the Kalman filter, the quantities (R) and (Q) are the covariances of the force vectors that cause the measurement signal noise covariance and state changes, respectively. When the Kalman filter is used to identify the matrix τ, the meaning of “measurement noise signal” is more generallyInterpretationIe, it must be interpreted as noise in the τ matrix,It is important to recognize. τline; queue; procession; paradeEach ofRatio of change in sensor response to change in actuator commandAsSince it is valid, “measurement noise” is a function of the change in actuator command.
[0034]
  Figure 3 shows the measured τ matrixofElement t₁₁Variability as a function of actuator command changeAnd thisByτ matrix noiseChanges in actuator commandDependIs shown. τ₁₁= ΔZ₁₁/ ΔU₁₁ BeI.e. t₁₁But,In response to changes in commands in actuator 1AgainstSensor 1 measurementIsRatio of change in response,I want you to remember that. To the actuatorA range offingerDecreeApply to systemdo itTo each change of actuator commandforA series of experiments was performed to measure the change in response that occurred. Figure 3 shows the change in command on the x-axis for a number of samples.When, Ratio of changes in response to command changes on the y-axis (ie τ matrix elements)And the relationshipThe results of plotting are shown. A hatched area 310 indicates a range of values. Small value of ΔUAgainstτ matrix value range is,For large values of ΔUAgainstThan the range of τ matrix values,Very widthofwideBecome a thing(Ie, τ matrix noise increases as ΔU decreases)Want to be. This relationship exists even if the noise associated with the measurement of Z is independent of U.
[0035]
  FIG. 3 shows that the covariance values (R) and (Q) need to be defined in order to control the noise. In many systems using Kalman filter, (R) and (Q) are kept constant.HaveThe When using a controller that uses a local model, that is, the “input” of the system is absoluteTypicalActuator forcenot,In the case of actuator force changes, it is more appropriate to see the quantities (R) and (Q) as a function of actuator command changes. Expressed appropriately, the information collected when the system is exposed to a large command is weighted more heavily in the estimation of τ than the information collected when the command change is small.
[0036]
In order to more accurately define the elements of the data matrix τ, the covariance (R) of the Kalman filter algorithm is defined as follows.
[0037]
[Equation 8]
R = (K_var/ | ΔU_i|)^N
Where K_varIs an empirically derived constant whose value is close to the value of ΔU that produces a higher response change than the noise present in the system. K_varThe value of is used as a weighting constant. Quantity | U_i| Is the magnitude of the control vector change, ie | ΔU_i| = SQRT (ΣΔU_i ²). N is an integer of any value, but is typically a real number greater than 1 and is established by testing. As described above, the effect of using (R) in the Kalman filter algorithm is that the change of the element of the data matrix τ is the actuatorof powerchangeΔU _i Is a function of
[0038]
In particular, the use of (R) in the form described above places emphasis on information gathered during large changes in actuator commands and not on information gathered during small changes in order to estimate τ. By doing so, the adverse effect of noise in the estimation of the τ matrix is greatly reduced.
[0039]
  FIG. 4 shows a flowchart of the steps for generating the control signal generated by the control circuit. The control signal includes actuator force data and data from a data matrix τ that has been repeatedly updated based on information previously or previously stored in the control circuit. This process starts by detecting vibration and generating a detected vibration signal, as shown in step 402. The detected vibration signal is sent to the control circuit as shown in step 404. Step 405 indicates that the detected vibration signal is processed. In step 406, the actuator force is calculated to minimize vibration detected by the vehicle. This calculation is based on the current values in the data matrix τ. In step 408, a signal for generating the actuator force calculated in step 406 is sent to the actuator. The actuator generates a vibration force that nullifies or detects a detected vibration of the vehicle. As shown in step 410, the ambient vibration is detected by the above-described sensor, for example. These sensors send the detected ambient vibration signal to the control circuit. In step 412, the control circuit receives the ambient vibration signal and uses the input to calculate an updated estimate in the data matrix τ. This update process utilizes Kalman filtration and covariance values (R). Covariance (R) has an adaptive feature that uses input data from previous compensation forces acting on the vehicle to determine the effectiveness of the force based on its magnitude and noise level. Use this covariance (R) to calculate a new estimate for the data matrix τCanTherefore, the second control signal is calculated to activate the second actuator force (ie, the vibration generator generates vibration in response to the second control signal). Typically, this control signal data is in a matrix formatAsAccumulated. When the control signal is generated by the control circuit, the control circuitthisA control signal based on the command matrix is sent to the actuator. These actuators receive the transmitted control signal and generate a compensation force to cancel the detected vibration, reducing ambient noise in the vehicle, as shown in step 413. This process is repeated as long as the system operates as shown in step 414.
[0040]
FIG. 5 shows a graph 50 of test data for a typical data matrix element τ, as identified by Kalman filter data. The size of the τ matrix element was plotted on the Y axis and the time was plotted on the X axis. FIG. 5 illustrates the advantages of the present invention by showing the size of the τ matrix element in response to changes in frequency input. A first frequency was introduced, a second frequency was entered, and then the system was returned to the first frequency to determine the τ matrix response (ie, a second frequency indicating a command was entered to enter the system for that command) The response was determined and how well the detection noise was reduced). During the first half of the test (time 0-1000 seconds) as indicated by line 512, the τ matrix elements varied. The second part of the experiment, starting at a time equal to 1000 seconds as shown by line 536, shows an improved τ matrix response.
[0041]
The portion of the graph shown as 514 shows the τ matrix elements identified during the first input to the actuator at the first frequency. The second frequency input was introduced at point 518. Point 520 represents the second input of the first frequency. As indicated by the 522 portion of the graph, the τ matrix elements varied substantially and did not return to the first frequency magnitude.
[0042]
At point 528, the present invention was implemented to determine the τ matrix response to changes in frequency. A first frequency was input at point 530 and a second frequency was input at point 531. At point 532, the first frequency was entered. Portion 534 indicates that the τ matrix element is invariant and has not changed substantially using the present invention.
[0043]
As shown in FIG. 5, the present invention generates an unchanging vibration load with reduced fluctuations. Relatively consistent operation with reduced fluctuations produces the desired vibration reduction characteristics and improves the environment inside the vehicle. The first part 512 of the controller repeats (time 0-1000 seconds) did not utilize the present invention and therefore varied significantly more than the second part 536 (time 1000-2000 seconds) utilized the present invention. . Therefore, when the τ matrix element is a function of covariance (R) as described here, the variation is significantly reduced. The improved tracking of the τ matrix element can be used to substantially reduce noise in the vehicle.
[0044]
FIG. 6 shows a graph 60 that is a plot of figure of merit data for τ matrix elements obtained with the test vehicle used in the test described with respect to FIG. The figure of merit was plotted on the Y axis and the time (seconds) was plotted on the X axis. The points shown in FIG. 6 correspond to the points in FIG. 5 except that 5 is changed to 6. During the first 1000 seconds of operation shown as line 612, testing was performed without implementing the present invention. A figure of merit was calculated for the first frequency operation at time zero, indicated by point 614. Point 618 indicates the figure of merit when the second frequency is introduced into the test. Point 620 indicates the figure of merit when the system operates at the first frequency. As shown by this graph, the figure of merit at point 622 substantially fluctuated.
[0045]
At point 628, a test was conducted using the improved on-line identification method of the present invention. Point 630 represents a figure of merit operating at the first frequency. A second frequency was input at point 631 and a first frequency was input at point 632. During the 634 portion, the figure of merit varied very slightly. The 636 portion (according to the invention) shows a significant reduction in the figure of merit variation compared to the 612 portion (not using the invention). As can be seen from FIG. 6, the improved filtration achieved by using a dynamic τ matrix that is a function of covariance (R) substantially eliminated the variation in figure of merit.
[0046]
The present invention is also applied to a flight simulator device. In such an embodiment, simulated vibrations and simulated actuator forces can be generated. The simulated data can be used to generate a data matrix. As a result, simulated vehicle motion can be obtained and stored in computer memory.
[Brief description of the drawings]
FIG. 1 shows a helicopter environment for the present invention.
FIG. 2 is a block diagram of components of the present invention.
FIG. 3 is a graph showing the relationship of τ matrix elements as a function of actuator command level changes.
FIG. 4 is a flow chart illustrating generation of a control signal that includes dynamically enabled and updated parameters.
FIG. 5 is a graph showing matrix elements of the dynamic data of the vehicle of the present invention.
FIG. 6 is a graph showing a figure of merit of the present invention.

Claims (9)

At least one sensor (110) adapted to be attached to the vehicle (122) and generating successive first and second ambient vibration signals corresponding to the first and second sensing conditions;
At least one actuator (114) adapted to be attached to the vehicle (122) and generating an actuator force applied to the vehicle (122);
A data module (147) coupled to the sensor (110) and the actuator (114) and storing data relating at least one actuator force to at least one sensor response to the actuator force;
Including a Kalman filter (140), continuously receiving first and second ambient vibration signals from the sensor (110), receiving data accumulated from the data module (147), and receiving the first ambient vibration signal; a control circuit sends the first control signal (118) to the actuator (114) in response to consist,
The control circuit (118) modifies the data in the data module (147) using a Kalman filtering algorithm and a covariance value that is a function of actuator force change ;
As attached to the vehicle (122), the control circuit (118) sends a second control signal that is a function of the second ambient vibration signal and the correction data of the data module (147). A device (10) that adapts to and improves the response of the vehicle to the detected condition. A vibration field in which the at least one actuator (114) reduces sensed ambient vibrations in the vehicle (122) in response to first and second control signals sent from the control circuit (118). The device according to claim 1, wherein: The apparatus according to claim 1, characterized in that the ambient vibration signal is stored as a sensor matrix, the actuator force is stored as an actuator matrix, and the data in the data module (147) is stored as a data matrix. . The apparatus of claim 3, wherein the elements of the data matrix are identified by a Kalman equation of the form:
[Expression 1]
τ _i = [τ ^T _i-1 + K _i (ΔZ _i ^T −ΔU _i ^T τ ^T _i-1 )] ^T
Here, K _i is a Kalman gain matrix, K _i = P _i ΔU _i R _i ⁻¹ , M _i = P _i−1 + Q _i−1 , P _i = M _i −M _i ΔU _i (ΔU _i ^T M _i ΔU _i + M _i ) ⁻¹ ΔU _i ^T M _i , ΔU is change in actuator force, ΔZ is change in sensor signal, R is covariance of noise measurement and function of actuator force, Q is covariance of force vector, T Is a transposed matrix, and τ is a matrix element. Covariance (R) is R = (K _var / | ΔU _i |) ^N , where K _var is an empirically derived weighting factor, ΔU is a change in actuator force, and N is a real number greater than 1. The device according to claim 4, characterized in that it is. The apparatus according to claim 1, characterized in that the data module (147) and the control circuit (118) are stored in a microprocessor. The device according to claim 1, characterized in that the vehicle (122) is a helicopter. Ambient vibration in the vehicle (122) is detected, an ambient vibration signal corresponding to the detected ambient vibration is generated, a constant value is divided by a change in actuator force, and a dynamic covariance corresponding to an integral power Processing the ambient vibration signal to generate a control signal that is a function, outputting the control signal to an actuator, and generating a vibration field as a function of the control signal to reduce ambient vibration A method of reducing vehicle (122) noise stored on a computer readable medium. A plurality of ambient vibration signals are accumulated in the sensor matrix, a plurality of control signals are accumulated in the data matrix, a plurality of actuator forces are accumulated in the actuator matrix, and the elements of the data matrix are corrected as a function of the actuator force. The method according to claim 8.

Images